Takes a field of study or practice and finds the few independent generators behind it, testing each set by whether it can regenerate the observed phenomena.
SKILL.md written in Chinese; this summary is our English description.
$ npx skills add lijigang/ljg-skills --skill ljg-rank -a claude-codeProject install by default; add -g for ~/.claude/skills/.
$ gh skill install lijigang/ljg-skills ljg-rank --agent claude-codeProject scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).
$ git clone --depth 1 https://github.com/lijigang/ljg-skills.git skills-src && mkdir -p .claude/skills && cp -r skills-src/skills/ljg-rank .claude/skills/ljg-rank && rm -rf skills-srcUse ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.
Claude Code skills documentation · loads skills from .claude/skills/
Install the "ljg-rank" agent skill from https://github.com/lijigang/ljg-skills/tree/master/skills/ljg-rank into .claude/skills/ljg-rank/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "ljg-rank", then confirm the skill loads.Claude Code copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$skill-installer install https://github.com/lijigang/ljg-skills/tree/master/skills/ljg-rankType this inside Codex. $skill-installer <name> installs a curated skill from openai/skills. The installer writes to $CODEX_HOME/skills (default ~/.codex/skills). Restart Codex if the skill does not show up.
$ npx skills add lijigang/ljg-skills --skill ljg-rank -a codexProject install goes to .agents/skills/; add -g for ~/.codex/skills/.
$ gh skill install lijigang/ljg-skills ljg-rank --agent codexProject scope by default (.agents/skills/); add --scope user for a personal install.
$ git clone --depth 1 https://github.com/lijigang/ljg-skills.git skills-src && mkdir -p .agents/skills && cp -r skills-src/skills/ljg-rank .agents/skills/ljg-rank && rm -rf skills-srcUse ~/.agents/skills/ instead of .agents/skills for a personal install.
Codex skills documentation · loads skills from .agents/skills/
Install the "ljg-rank" agent skill from https://github.com/lijigang/ljg-skills/tree/master/skills/ljg-rank into .agents/skills/ljg-rank/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "ljg-rank", then confirm the skill loads.Codex copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$ npx skills add lijigang/ljg-skills --skill ljg-rank -a cursorProject install goes to .agents/skills/; add -g for ~/.cursor/skills/.
$ gh skill install lijigang/ljg-skills ljg-rank --agent cursorProject scope by default (.agents/skills/); add --scope user for a personal install.
$ git clone --depth 1 https://github.com/lijigang/ljg-skills.git skills-src && mkdir -p .cursor/skills && cp -r skills-src/skills/ljg-rank .cursor/skills/ljg-rank && rm -rf skills-srcUse ~/.cursor/skills/ instead of .cursor/skills for a personal install.
Cursor skills documentation · loads skills from .cursor/skills/, .agents/skills/, .claude/skills/, .codex/skills/
Install the "ljg-rank" agent skill from https://github.com/lijigang/ljg-skills/tree/master/skills/ljg-rank into .cursor/skills/ljg-rank/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "ljg-rank", then confirm the skill loads.Cursor copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$ gemini skills install https://github.com/lijigang/ljg-skills.git --path skills/ljg-rank--scope user (default) or --scope workspace; --path is the subfolder of the repo that holds the skill; --consent skips the security confirmation prompt.
$ npx skills add lijigang/ljg-skills --skill ljg-rank -a gemini-cliProject install goes to .agents/skills/; add -g for ~/.gemini/skills/.
$ gh skill install lijigang/ljg-skills ljg-rank --agent gemini-cliProject scope by default (.agents/skills/); add --scope user for a personal install.
$ git clone --depth 1 https://github.com/lijigang/ljg-skills.git skills-src && mkdir -p .gemini/skills && cp -r skills-src/skills/ljg-rank .gemini/skills/ljg-rank && rm -rf skills-srcUse ~/.gemini/skills/ instead of .gemini/skills for a personal install, then run /skills reload.
Gemini CLI skills documentation · loads skills from .gemini/skills/, .agents/skills/
Install the "ljg-rank" agent skill from https://github.com/lijigang/ljg-skills/tree/master/skills/ljg-rank into .gemini/skills/ljg-rank/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "ljg-rank", then confirm the skill loads.Gemini CLI copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$ gh skill install lijigang/ljg-skills ljg-rankInstalls for Copilot at project scope by default; add --scope user for a personal install. Preview a skill first with gh skill preview. Needs GitHub CLI 2.90.0 or later (public preview).
$ npx skills add lijigang/ljg-skills --skill ljg-rank -a github-copilotProject install goes to .agents/skills/; add -g for ~/.copilot/skills/.
$ git clone --depth 1 https://github.com/lijigang/ljg-skills.git skills-src && mkdir -p .github/skills && cp -r skills-src/skills/ljg-rank .github/skills/ljg-rank && rm -rf skills-srcUse ~/.copilot/skills/ instead of .github/skills for a personal install. Commit .github/skills so cloud agent and code review can use it.
GitHub Copilot skills documentation · loads skills from .github/skills/, .claude/skills/, .agents/skills/
Install the "ljg-rank" agent skill from https://github.com/lijigang/ljg-skills/tree/master/skills/ljg-rank into .github/skills/ljg-rank/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "ljg-rank", then confirm the skill loads.GitHub Copilot copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$ npx skills add lijigang/ljg-skills --skill ljg-rank -a opencodeOpenCode documents no install command of its own. Project install goes to .agents/skills/; add -g for ~/.config/opencode/skills/.
$ gh skill install lijigang/ljg-skills ljg-rank --agent opencodeProject scope by default (.agents/skills/); add --scope user for a personal install.
$ git clone --depth 1 https://github.com/lijigang/ljg-skills.git skills-src && mkdir -p .opencode/skills && cp -r skills-src/skills/ljg-rank .opencode/skills/ljg-rank && rm -rf skills-srcUse ~/.config/opencode/skills/ instead of .opencode/skills for a personal install.
OpenCode skills documentation · loads skills from .opencode/skills/, .claude/skills/, .agents/skills/
Install the "ljg-rank" agent skill from https://github.com/lijigang/ljg-skills/tree/master/skills/ljg-rank into .opencode/skills/ljg-rank/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "ljg-rank", then confirm the skill loads.OpenCode copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
ljg-rankTakes a field of study or practice and finds the few independent generators behind it, testing each set by whether it can regenerate the observed phenomena.
Given a domain, the skill looks for its rank: the smallest set of independent generators from which every phenomenon in the field can be reproduced. It first surfaces the domain's basic assumptions, the things that rest on belief and are not allowed to be questioned, and writes them down as a ceiling, then digs downward toward the root rank.
The digging follows a seven-step process that stays out of the final text: lay out 10+ representative phenomena, list candidate generators, sink one layer at a time, merge candidates with the same source, cut, regenerate the phenomena, and run a prediction test plus a hard-to-vary test taken from David Deutsch. Five internal criteria then check generativity, minimality, independence, reach and resistance to variation. A references/template.org file is included, and the SKILL.md is written in Chinese.
10 steps, taken from the step headings in SKILL.md.
Read from SKILL.md and the folder at commit 9e75497. It shows what the files ask for, not the result of running them.
Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.
From allowed-tools in the SKILL.md frontmatter.
No scripts in the folder and no shell commands in SKILL.md.
From the folder's file list and the shell code blocks in SKILL.md.
No URLs in SKILL.md.
From URLs in SKILL.md, links to its own repository left out.
Names no API keys, tokens, secrets or passwords.
From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.
Rank Reduction Engine loads about 3.2k tokens when it runs, and up to ~3.2k if it reads all its reference files. Until then it costs about 48 tokens; SKILL.md has 506 words of instructions outside code blocks.
Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.
The automated check found no risky patterns in SKILL.md.
Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.
The full file from lijigang/ljg-skills at commit 9e75497, republished under its MIT licence (© lijigang). 506 words, ~3,218 tokens.
.claude/skills/ljg-rank/SKILL.md (or your agent's skills folder). This skill also uses 1 other file; get the full folder from GitHub.输入一个领域，输出它的秩。
秩不是"关键要素",不是"核心原则",不是"总结要点"。
秩是这么个东西:这个领域里真正独立的生成器,究竟有几根?拿这几根,能不能把全部现象一个个倒回来?能,才算找到。
但能倒回来只是底线。Deutsch 在《无穷的开始》里立的尺——好解释要过两关:
坏解释能用十种说法糊过去——它根本不是在解释,是用模糊把面铺得很广。好解释只有这一种说法能把所有现象对上——动一根就塌。
找秩,找的就是这种"动一根就塌"的好解释。
挖之前,先抬头看一眼这个领域立在什么上面——它的基本假设是哪几条?
一套理论,按亚里士多德的规矩,总要有几条不证自明的起点。基督教就立在三条上:
这三条不许问"为什么"——它们靠信,不靠证。动摇任何一条,整套教义就塌。
每个领域都有这种东西。物理学里是"自然律在时空中稳定"——你没法用实验去证,实验本身就预设了它。经济学里是"人会按自己的偏好排序选择"——这不是观察出来的结论,是入场的门票。哲学里是"语言能指称世界"——分析哲学整个家都建在这块地基上。
降秩之前,先把这几条挖出来,明明白白写在纸上。
操作问句两条,对着领域问:
答出来的几条,就是这个领域的基本假设。
假设和 rank,一句话分清:
两个方向,一上一下。地板没找到,鬼打墙;天花板没看清,容易把信仰当真理说出去——后者更危险,因为听上去也像真理。
写进文章里的方式:开头一两句,把基本假设交代清楚。不展开论证,不解释"为什么是这几条"——它们的位置就是不证。然后才开始往下挖 rank。
判据是事后才能验的事,挖 rank 的力气全在过程里。判据再严,也只能挑出伪根——前提是工序走到了底。
七步,心里走完,不写进文章:
铺现象——领域里 10+ 个有代表性的现象,光铺,不解释。要的是给 rank 留一份可反生成的检验靶。
列候选——每个现象问一句"为什么会这样",候选生成器全贴出来。这一步必然杂,伪根混在里头。
递归下沉——这是穿透的命门,也是最容易做错的一步。错法是想一铲子挖到底:从现象直接猜"最根本的是什么"。站在现象层看不见地基,只看得见脚下这一层,猜出来的底十有八九悬空。
对的做法是一层一层沉下去。先只找第一层生成器——现象底下紧挨着的、看得见够得着的那层。找齐了,把这一层当新地面,人站上去。 站稳了再问一遍"这层底下还有没有更基本的"——这时候第二层才露出来,它从现象层根本无从发问。第二层找齐,再当新地面,再站上去,再往下看。一级一级递归,直到站上某一层,底下再问不出更基本的——那层就是 root rank。
这正是黑箱往下开一层的笨功夫:当前这层是能监测的全部,看清了、站稳了,下一层才打开。想跳级是贪快,跳一层,root rank 就虚一层。
比如降秩"创业",第一铲挖出愿景、执行、团队、市场——看着都对,其实只是表面的"关键要素",停在这儿就是没下沉。把这四个当新地面站上去,再问:愿景从哪来?执行靠什么撑?这才下到第二层。一层层站到追不动,才摸到 root rank。
合并同源——两根候选其实是同一个生成器在两面上露脸,合掉。判别一句话:去掉一个,另一个还会不会自己冒出来?
砍——拿掉一根,剩下的能不能反生成全部现象?砍得掉的扔,砍不掉的留。这一步把判据里的"最小性"落到地。
反生成——剩下的几根,从头把现象清单走一遍,看能不能逐条复现。复现不了的那条,指向少了一根。
预测 + 变更双测——两道闸一起过:
七步走完,几根候选才够格放上判据台。
工序心里走透,但别写进文章——和判据一样,是内功不是叙事钩子。读者看不到挖的过程,但感觉得到背后是凿到底的。
工序走完之后,这五条是事后再过一遍的闸,同样不是写出来给人看的。
前三条管骨架(这几根的形状对不对):
后两条管解释力(够不够称得上好解释)—— Deutsch《无穷的开始》: 4. 解释力(reach)——能推出清单外的现象,而且现实里真有 5. 难以变更(hard to vary)——改一根的任何一个细节,预测就错(管的是"细节"——不能动)
最狠的是第五条。最小性管"数量"——不能拿走;hard to vary 管"细节"——不能动。一个生成器,要是"换种说法也通""调一下也成立",它就是松的伪根,预测力是装出来的——靠的是模糊把面铺广,不是靠精准。坏解释怎么改都还能用;好解释一动就塌——这是 Deutsch 在《无穷的开始》里立的核心分判。
过不了一条,推倒重来——回到工序第 3 步接着追。第五条过不了,往往不是少了一根,而是某一根写得太松——逼自己改到"这根只能这么说,改不动"为止。
判据在心里走结实,但别落到纸上——不要出现"砍一刀看看""拆一根试试"这种验证动作。这些会打断读者从"乱"到"简"的那一口气。读者看不到验证过程,但感觉得到底下是硬的。
判据过完,几根线浮出来了。先别急着写——分两层看它们长成什么样。
第一层:找 root rank(根骨架)
root rank 是这个领域真正的根骨架,也是递归下沉的终点——一层层站下去,站到再也站不下去的那层。不预设几根,不预设根与根之间什么关系,就找。它可能长成:
读完 root rank,读者手里拿到的是世界观——这领域的骨架,一种看世界的新方式。
形状反问 · 反坍缩闸(找完几根线,下笔前强制自检)
九种形状,逐一过一遍——
AI 一上手就爱想到"钻井剖面"——这是懒惰锚点。任何概念都能拆三层(表层/中层/底层),拆完看着齐整,但很多东西的真实结构根本不是垂直往下钻的。注意:递归下沉挖出的纵深,和这里的懒惰三层不是一回事——下沉是挣来的,真站到每层、验过它生成上一层,垂直结构才作数;没下沉就先摆三层,才叫懒惰。而且纵向有几层、和每层内部那几根是什么形状(并排/拉扯/互推),要分开算。
强制反问:
把另外 8 种逐一排除掉,才能落到"钻井剖面"。这道排除不能跳——跳过的代价是 root rank 失真,把先后关系、反馈关系、张力关系一股脑压成嵌套。
第二层:能不能叠一台操作坐标系
root rank 找完了,再问一句:里头有没有哪一根线,能展开成几个独立可调的状态参数?
举例:Fritz 的 root rank 第三根是"火箭几何 vs 钟摆几何"。火箭几何要"愿景"和"现状"两个点;把这两点的清晰度、诚实度当成可调的轴,就长出一个二维坐标系。坐标系不是 root rank 的替身,是它在某根线上长出来的孩子。
能展开的话,坐标系就叠在 root rank 之上。读者读完坐标系,多拿到一样东西:操作仪——任何具体情境往框里一摆,就能看出落在哪个象限、哪一层。
两层不打架——机制讲"是什么",坐标系讲"在哪儿",答的是两个问题,可以一起给。
什么时候叠,什么时候不叠:
能叠坐标系,秩就升了一档——读者走出来,手里既有世界观,也有一台操作仪。
取景框反问 · 反坍缩闸(展开第二层之前,强制自检)
确认要叠之前,先问一句:我要展开的这根线,真的是 两根独立可滑动的轴 吗?
AI 的另一个懒惰锚点是 2x2 矩阵。咨询和战略框架的训练数据里 2x2 极多,任何两个变量都能切四象限,切完看着像在分析。但很多 root rank 不是两个独立维度——是单一光谱、是反馈环、是链式。
强制反问:
把光谱/环路/链式/树形挨个排除掉,才能确认 2x2。
下笔前,默念三遍:「一个没读过英文的中国人,会这么说话吗?」
每写完一段,再默念一遍。答"不会"——别改词,整段推倒重想。
最容易踩的英语化句式(踩到就重写整段,不是抠字眼):
| 英语化 | 中文母语 |
|---|---|
| 它建立在前一根上 | 前一根托着它 |
| 它有内在方向 | 它自己往那儿走 |
| 这一区分回答的是 X,不是 Y | 这问的是 X,不是 Y |
| 这意味着... | 也就是说... / 这就是为什么... |
| 在...的过程中 | ...的时候 / 在... |
| 对...而言 | 对...来说 |
| 是...的(强调句) | 直接说 |
| 进行 + 名词 | 换成动词 |
| 名词化抽象("X 性"、"Y 化") | 换成具体的动词或物件 |
要改的不是词,是表达的取景框——从英语 critical thinking 的句式,切到中文白话作者的笔调(汪曾祺、王小波、阿城、李娟那一路)。
事后扫词只能修个表面,治不了根。根在"想的时候就是英文"。改写时,把英文先撂下,蹲在中文里重新想一遍这件事该怎么说出来。
写完默念出声,听耳朵卡不卡。卡的地方,整段重写,不是修补。
一气呵成的散文。
带读者走一段路——从"这个领域看着挺乱"走到"原来就这几根线在牵"。不分章节,不加小标题,让推理自己推着往前走。
三条要求:
两段式(如果叠了坐标系):
两段中间不要硬切。让推理顺过去——上半最后一句最好留个引子,像"这套机制落到具体情境里怎么看?",把读者顺到下半。
光说"这几根线"不够——要把这几根之间的关系画出来。
但画法不止一种。root rank 长成什么样,取景框就该长成什么样。下笔前先问一句:这个 rank 的关系,本身是什么形状? 答完再选画法。形式跟着骨架走,不是反过来。
九种取景框家族——这一张表是骨架到画法的总索引:
| Root rank 形态 | 关系本质 | 取景框 |
|---|---|---|
| 几根并排独立、可滑动 | 正交 | 二轴 / 多轴坐标系 |
| 一层托一层(最深一根是元命题) | 嵌套穿透 | 钻井剖面 |
| 一根线两端拉扯 | 张力对立 | 光谱 / 滑标 |
| 互相正负推动 | 反馈循环 | 环路图(标 +/-) |
| 一段接一段 | 阶段递进 | 链式 / 台阶 |
| 一根分多根,多根再分 | 层级分叉 | 树形图 |
| 多对多互勾 | 耦合网络 | 网状图 |
| 涨涨落落、节奏交替 | 振荡 | 波形 / 振荡曲线 |
| 多维分类(如抽象度 x 远近度 x 时间) | 多维分类 | N 轴 / 多切片 |
硬约束：只用纯 ASCII 字符。禁用任何 Unicode 符号（包括箭头 → ← ↑ ↓、方框 ┌─┐└┘├┤│、圆点 • ◆ ●、粗体 ▶ ◀ 等）。
允许字符集：字母、数字、中文汉字、空格，以及 - = | + * / \ < > ^ v [ ] ( ) { } . , : ; _ #。
对照表（左禁用 / 右替换）：
┌ ┐ └ ┘ ├ ┤ ┬ ┴ ┼ -> +─ ━ -> -（或 = 表粗线）│ ┃ -> |→ ▶ -> ->← ◀ -> <-↑ -> ^ 或 |^↓ -> v 或 |v◀─▶ -> <-> 或 <--->● -> * 或 o× -> x下面九种范例——每种配一张最小可用的 ASCII 图 + 一行演示要求。挑哪一种，跟前面工序穿透出来的 root rank 形状走。
适合:root rank 的几根线正交、独立、可滑动。Fritz 火箭几何(愿景清晰度 x 现状诚实度)就是这种。
演示要求:图里必须标几个具体现象的坐标——比如"拥抱"落在哪、"问候"落在哪、"熵的局部逆转"落在哪。读者一眼看见这框怎么用,才学得会。空有坐标轴没有演示点,框就是空摆设。
坐标系分两档,决定每个象限里画什么:
属性型坐标系——两根轴是分类或状态属性(比如轻-重 x 情感-理性、外向-内向 x 直觉-感知)。每个象限只是个分类标签,格子里没什么动力学要发生。按上面的要求做就行——四角标演示坐标和例子,格子里不再画。
动力学型坐标系——两根轴是动力学维度,每个象限里的系统会演化出独特的时间行为。像势函数(保守-耗散 x 单-多极小)、相变(温度-压强)、博弈(零和-非零和 x 一次-重复)。这种坐标系扁平地"贴标签"是浪费——每个象限都该升级成带矩形边框的微型 ASCII 画,把"这格里系统怎么动"裱在框里直接画出来。
一句话判别:这格里的系统,有没有一段画得出来的动作? 有——升级为动力学画面;没有——退回属性标签。
动力学画面怎么画:
+ - | 围,像一幅装裱过的小 ASCII 画)势函数坐标系范例：
单极小
|^
|
|
+--------------------------+ | +--------------------------+
| | | | |
| \________/ | | | \________/ |
| o <-> | | | o . |
| | | | |
| 唯一谷 球永恒摆动 | | | 唯一谷 球减速落底 |
| | | | |
| 保守 + 单极小 | | | 耗散 + 单极小 |
| [钟摆 / 谐振子] | | | [凸优化 / SVM] |
+--------------------------+ | +--------------------------+
|
保守 <-------------------------+-------------------------> 耗散
(能量守恒) | (能量被环境吸走)
|
+--------------------------+ | +--------------------------+
| | | | |
| \_/\_/\_/\_/\_/ | | | \_/\_/\_/\_/\_/ |
| <-o o o-> | | | o . o |
| | | | |
| 多谷 能量越脊穿梭 | | | 多谷 球落各谷不出 |
| | | | |
| 保守 + 多极小 | | | 耗散 + 多极小 |
| [三体 / 混沌系统] | | | [神经网络 / 蛋白质折叠] |
+--------------------------+ | +--------------------------+
|
|v
多极小读者一眼就能看见:每格里"地形 + 球的命运"裱成一幅画,系统怎么演化直接可见。这是分类标签做不到的。
适合:root rank 一层托一层,最深一根是元命题。读者顺着垂直方向往下穿。
+------------------------------------+
| 表层 现象 |
| [用户每天打开 App] |
+------------------------------------+
|
v
+------------------------------------+
| 中层 机制 |
| [推送 + 反馈 + 习惯回路] |
+------------------------------------+
|
v
+------------------------------------+
| 底层 元命题 |
| [多巴胺间歇强化] |
+------------------------------------+演示要求:每一层都要标具体例子,让读者顺着垂直方向看见"现象 -> 机制 -> 元命题"的穿透路径。最深一层是元命题,别留空——留空就是没钻到底。
适合:root rank 是一根线两端在拉,案例落在中间某点。
极端A <===================o===================> 极端B
[偏A 的特征] ^ [偏B 的特征]
|
案例 X 落在这演示要求:两端的特征要标出来,中间某点至少放一个具体案例的位置。光有两端没有案例点 = 空摆设——读者得看见这把尺子是怎么量的。
适合:root rank 是几根互相推的环,正负方向必须画出来。
A --(+)--> B
^ |
| |
(-) (+)
| |
| v
D <--(+)-- C演示要求:每条箭头要标 + 或 -;正环(reinforcing,自我加强)和负环(balancing,自我抑制)的总效应在图旁边注一句——不写正负的环路图等于没画。
适合:root rank 是阶段递进,时间或逻辑上一段接一段。
链式横排：
A ---> B ---> C ---> D
[阶段名] [阶段名] [阶段名] [阶段名]台阶纵排（节节升）：
+----+
| D |
+----+
| C |
+----+
| B |
+----+
| A |
+----+演示要求:每段要命名,还要标转换条件(什么把 A 推到 B,什么把 C 升到 D)。没有转换条件的链 = 罗列,不是机制。
适合:root rank 一根分多根,多根再分。
root rank
/ | \
A B C
/ \ | /|\
a b d x y z演示要求:每个分叉点要命名分类原则(按什么把上一层切成下一层)——不命名分类原则的树是糊的。
适合:root rank 多对多互勾,没有层级。
A --------- B
|\ /|
| \ / |
| \ / |
| \ / |
| \ / |
| X |
| / \ |
| / \ |
| / \ |
| / \ |
D --------- C演示要求:每条边要有方向(若有方向)和强度(粗线 = / 细线 -),关键节点要标。一片纯连线不分主次 = 网,但不是骨架。
适合:root rank 是涨涨落落、节奏交替(如季节、周期、繁荣-萧条、专注-涣散)。
/\ /\ /\
/ \ / \ / \
/ \ / \ / \
/ \____/ \____/ \____
^ ^
谷 峰
[低谷期表现] [高峰期表现]演示要求:至少标一对谷-峰具体例子,让读者看见周期里不同位置的状态——不标位置的波形等于装饰线。
适合:超过两轴的多维分类(比如 Y 抽象度 x X 远近度 x Z 时间)。3D 难画清,画两两 2D 切片:
切片一: 抽象度 x 远近度 切片二: 抽象度 x 时间
高抽象 高抽象
^ ^
| |
| |
近 <-----+-----> 远 过去 <---+---> 现在
| |
| |
v v
低抽象 低抽象演示要求:每张切片至少标一个案例落在哪个象限;两两切片画全了才算完整。漏一对 = 留盲区。
叠加式(root rank + 取景框)画两张图:上张是 root rank 的骨架(用上面九种里选一种),下张是叠在某根线上的取景框(标好演示坐标 / 演示点 / 演示路径)。两张图之间用一条线牵——比如"此处展开 ->"或"在这一格里再放大"——告诉读者第二张嵌在第一张的哪一格里。读者一眼看见骨架,又看见骨架某一格里还藏着一台取景仪,世界观和操作仪同时入脑。
不设死板模板。要见骨——标出生成器的名字、关系的方向、正负的影响。读者看图一眼,骨架入脑。
date +%Y%m%dT%H%M%S 和 date "+%Y-%m-%d %a %H:%M"~/Documents/notes/{时间戳}--{领域}的秩__rank.org© lijigang, MIT. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file
SKILL.md and 1 other file (references) in skills/ljg-rank of lijigang/ljg-skills.
Open the folder on GitHubat commit 9e75497
Rank Reduction Engine next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.
| Skill | Stars | Used in | Tokens | Auto-check | Licence | Repo updated |
|---|---|---|---|---|---|---|
| Rank Reduction Engine this skilllijigang/ljg-skills | 7.5k | — | ~3.2k | Automated safety check: Pass | MIT | |
| Paper Opportunity Radartamdogood/builder-essential-skills | 221 | — | ~3.1k | Automated safety check: Pass | MIT | |
| PrimekgK-Dense-AI/scientific-agent-skills | 48k | 1 repos | ~2.4k | Automated safety check: Pass | MIT | |
| Ccf Idea Optimizermikubaka88/CCFA-Skills | 3k | — | ~1.9k | Automated safety check: Pass | MIT | |
| Hypothesis Generationspacering-net/codeg | 3.9k | 14 repos | ~3.6k | Automated safety check: Notes | MIT | |
| Nature Paper CardYuan1z0825/nature-skills | 47k | 2 repos | ~2.1k | Automated safety check: Pass | Apache-2.0 |
tamdogood/builder-essential-skills
Run a cumulative daily or retrospective sweep of research papers on a chosen topic, audit their claims, methods, integrity signals, and independent support, then identify overlooked but feasible…
K-Dense-AI/scientific-agent-skills
Queries a pinned Precision Medicine Knowledge Graph (PrimeKG) CSV for typed gene, drug, disease, and phenotype nodes, direct associations, disease context, and one- or two-hop paths.
mikubaka88/CCFA-Skills
Develop and optimize rough CCF research ideas into problems, insights, mechanisms, and evidence plans.
spacering-net/codeg
Structured hypothesis formulation from observations. An agent skill from spacering-net/codeg.
Yuan1z0825/nature-skills
Builds a structured deep-reading card for one scientific paper, covering methods, how experiments support claims, limitations and research ideas, with a script to prepare the source.
K-Dense-AI/claude-scientific-writer
Formulate evidence-bounded scientific questions, candidate hypotheses, rival explanations, causal or associational claims, discriminating predictions, measurements, and preregistration-ready…
lijigang/ljg-skills
Explains a whole book to someone who has not read it, keeping its specific content and showing how its threads connect, and saves the result as an Org note.
lijigang/ljg-skills
Explain research papers to readers without a specialist background: what the paper studies, what the authors contribute, how the findings follow, and what the evidence does not establish.
lijigang/ljg-skills
Turns text, URLs or local files into tall PNG cards through HTML typography, with four modes: long reading card, full-text layout, comic and whiteboard.
lijigang/ljg-skills
Turns a named classical Chinese chapter, such as one from the Tao Te Ching or the Analects, into a single annotated PNG image with notes and commentary.
lijigang/ljg-skills
Finds the handful of constraints that truly define a domain, role, product or debate, grades each by hardness, and explains the behavior those constraints produce.
lijigang/ljg-skills
Builds single-file offline HTML talk decks from Org or Markdown outlines, with faithful layout or editorial condensing and keyboard navigation.
Categories
Takes a field of study or practice and finds the few independent generators behind it, testing each set by whether it can regenerate the observed phenomena. Given a domain, the skill looks for its rank: the smallest set of independent generators from which every phenomenon in the field can be reproduced. It first surfaces the domain's basic assumptions, the things that rest on belief and are not allowed to be questioned, and writes them down as a ceiling, then digs downward toward the root rank.
Rank Reduction Engine fits situations like: decomposing a field into its few independent driving forces; testing whether a proposed set of causes can regenerate all known phenomena; surfacing the unquestioned assumptions a discipline rests on; preparing an essay on what really holds a domain up.
Run `npx skills add lijigang/ljg-skills --skill ljg-rank -a claude-code`. Or copy the skill folder (skills/ljg-rank in lijigang/ljg-skills) into .claude/skills/ljg-rank in your project. Claude Code loads it when a task matches its description.
Run `npx skills add lijigang/ljg-skills --skill ljg-rank -a codex`. Or copy the skill folder (skills/ljg-rank in lijigang/ljg-skills) into .agents/skills/ljg-rank in your project. Codex loads it when a task matches its description.
Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add lijigang/ljg-skills --skill ljg-rank -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/ljg-rank, .gemini/skills/ljg-rank, .github/skills/ljg-rank and .opencode/skills/ljg-rank in your project.
SKILL.md names no scripts, command-line tools or credentials: Rank Reduction Engine is instructions for the agent only.
SKILL.md contains no URLs. Any network use would come from the scripts or tools the agent runs. This is read from the text; nothing was executed.
Our automated static check of SKILL.md found no risky patterns, such as piping downloads into a shell, reading credential files or hidden Unicode. It is not a guarantee. Review the folder before installing.
Rank Reduction Engine is published under the MIT licence (the repository's licence). It allows redistribution, so the full SKILL.md is shown on this page.
About 3.2k tokens (SKILL.md is roughly 13k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full. Its references folder adds about 31 tokens, read only when the agent opens those files.
Skills that share tags, products or a category with Rank Reduction Engine: Paper Opportunity Radar (tamdogood/builder-essential-skills, 221 stars), Primekg (K-Dense-AI/scientific-agent-skills, 48k stars), Ccf Idea Optimizer (mikubaka88/CCFA-Skills, 3k stars) and Hypothesis Generation (spacering-net/codeg, 3.9k stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.
lijigang (a GitHub user) maintains it in lijigang/ljg-skills, which has 7,481 GitHub stars. The repository holds 26 skills in this directory. The repository was last updated on October 8, 2026.
Source: lijigang/ljg-skills on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.
